def derivativeFunction(x, N, N_im, sz, dims, dimOpt, dimLenOpt, lam1, lam2, data, k, strtag, ph,
kern, dirWeight=0.1, dirs=None, dirInfo=[None,None,None,None], nmins=0, wavelet="db4", mode="per", a=10.):
'''
This is the function that we're going to be optimizing via the scipy optimization pack. This is the function that represents Compressed Sensing
'''
disp = 0
gTV = 0
gXFM = 0
x.shape = N
if len(N) > 2:
x0 = np.zeros(N_im)
for i in xrange(N[0]):
x0[i,:,:] = tf.iwt(x[i,:,:],wavelet,mode,dims,dimOpt,dimLenOpt)
else:
x0 = tf.iwt(x,wavelet,mode,dims,dimOpt,dimLenOpt)
gdc = grads.gDataCons(x0,N_im,ph,data,k,sz)
#import pdb; pdb.set_trace()
if lam1 > 1e-6:
gtv = grads.gTV(x0,N_im,strtag,kern,dirWeight,dirs,nmins,dirInfo=dirInfo,a=a)
gDataCons = np.zeros(N)
gTV = np.zeros(N)
gXFM = np.zeros(N)
for i in xrange(N[0]):
gDataCons[i,:,:] = tf.wt(gdc[i,:,:],wavelet,mode,dims,dimOpt,dimLenOpt)[0]
if lam1 > 1e-6:
gTV[i,:,:] = tf.wt(gtv[i,:,:],wavelet,mode,dims,dimOpt,dimLenOpt)[0] # Calculate the TV gradient
if lam2 > 1e-6:
gXFM[i,:,:] = grads.gXFM(x[i,:,:],a=a)
x.shape = (x.size,)
return (gDataCons + lam1*gTV + lam2*gXFM).flatten() # Export the flattened array
def derivativeFunction(x,
N,
N_im,
sz,
dims,
dimOpt,
dimLenOpt,
lam1,
lam2,
data,
k,
strtag,
ph,
kern,
dirWeight=0.1,
dirs=None,
dirInfo=[None, None, None, None],
nmins=0,
wavelet="db4",
mode="per",
a=10.):
'''
This is the function that we're going to be optimizing via the scipy optimization pack. This is the function that represents Compressed Sensing
'''
disp = 0
gTV = 0
gXFM = 0
x.shape = N
if len(N) > 2:
x0 = np.zeros(N_im)
for i in xrange(N[0]):
x0[i, :, :] = tf.iwt(x[i, :, :], wavelet, mode, dims, dimOpt,
dimLenOpt)
else:
x0 = tf.iwt(x, wavelet, mode, dims, dimOpt, dimLenOpt)
gdc = grads.gDataCons(x0, N_im, ph, data, k, sz)
#import pdb; pdb.set_trace()
if lam1 > 1e-6:
gtv = grads.gTV(x0,
N_im,
strtag,
kern,
dirWeight,
dirs,
nmins,
dirInfo=dirInfo,
a=a)
gDataCons = np.zeros(N)
gTV = np.zeros(N)
gXFM = np.zeros(N)
for i in xrange(N[0]):
gDataCons[i, :, :] = tf.wt(gdc[i, :, :], wavelet, mode, dims, dimOpt,
dimLenOpt)[0]
if lam1 > 1e-6:
gTV[i, :, :] = tf.wt(gtv[i, :, :], wavelet, mode, dims, dimOpt,
dimLenOpt)[0] # Calculate the TV gradient
if lam2 > 1e-6:
gXFM[i, :, :] = grads.gXFM(x[i, :, :], a=a)
x.shape = (x.size, )
return (gDataCons + lam1 * gTV +
lam2 * gXFM).flatten() # Export the flattened array